I wanted to run a couple of sets of extension speakers from my stereo, into the workshop and electronics lab. I decided to use seperate amplifiers at the remote locations, so that the power and volume controls were conveniently located, and to avoid excessively long speaker cables. However, connecting the amplifiers directly together with shielded audio cable at line level is a recipe for earth loops, and hum on the signal. I have experienced this problem with just a computer and amplifier on opposite sides of the room, and it could potentially be much worse with cables running right through the house.
One possible solution would be to use a balanced audio connection, with appropriate driver and receiver circuits. However, I decided instead to use optocouplers in the signal path at each of the slave amplifiers, to completely remove the possibility of earth loops. A circuit at the main amplifier would drive a twisted pair current loop, which could be wired in Cat. 5 cable. The optocoupler LEDs for each of the remote stations would simply be connected in series.
Simply biasing the LED and driving it with the audio signal voltage could result in quite poor linearity and high distortion. There are a couple of ways around this. Special 'linear' optocouplers are made with a second photo-transistor, which can be used in a feedback loop to regulate the light output of the LED. However, this application did not justify the complexity and expense that this approach would entail. I did not need full 'hi-fi' performance from the circuit, as it was mainly to be used for casual listening.
A simpler approach is to drive the LED with a current proportional to the signal voltage. This will remove most of the nonlinearities of the LED, especially if a high standing bias current is used relative to the magnitude of the audio signal.
The circuit shown above was developed towards this end. The LED current flows through R2, which functions as a current-to-voltage converter. The current signal is compared to the input audio signal by the op-amp, and the LED is driven so as to make the current flowing proportional to the input signal. RV1 sets the bias potential, which is buffered by the second op-amp section. The circuit operates from the +/-12V rails in the amplifier. The entire circuit is duplicated for stereo operation. While it might be possible to share the bias supply across both channels, it is just as easy to build a second one, as there is a spare op-amp section anyway.
The receiver circuit is quite simple - just three components! This is an advantage if it is replicated at multiple receiving stations. The input is completely floating, so the stations are simply connected together in a series loop. (Note, however, that the wiring should always be run as a twisted pair to reduce noise pickup.). I ended up using standard 4N25 optocouplers, which are cheap and readily available - I already had quite a few that I had salvaged from old equipment. Although these are generally regarded as fairly slow devices, the frequency response is quite adequate provided the bias is set appropriately. A trimpot is used to set the bias current, and a capacitor couples the audio into the amplifier input. The power supply is taken from the amplifier, which in this case was a single ended +12V. The circuit could be used from a dual supply, or a different voltage, although a different value trimpot may be required.
Circuit board under construction
Backlighting the board to check for shorts
I decided to try a new construction technique for this circuit. Normally I would probably use perf board for a one-off circuit like this, but I can't get hold of this easily anymore. Instead of going to the trouble of designing and fabricating a PCB in the usual manner, I decided to try mechanically etching a board instead. This allows the board to be designed 'on-the-fly'. I used a 1mm carbide burr in a high speed rotary tool to divide the blank copper board into tracks. It helps to backlight the board to check that the copper lands have been fully isolated. The components were soldered onto the copper side to avoid the need to drill holes, with some of them being glued in place for extra stability.
The mechanically etched board turned out to be a quick way to get the circuit up and running. (I had previously tested the circuit on a breadboard). Unfortunately, when powered up, one channel did not function correctly - this was traced to a short on the board. This is obviously something to watch for, as it does detract somewhat from the 'rapid prototyping' nature of this method!
The receiver circuit is too simple to require a PCB, and was built on the back of a DIN plug installed in the slave amplifier.
Before installing the circuit, some measurements of its performance were made. These are all fairly rough measurements, but should give an idea of the circuit's capabilities.
The transfer characteristics are very much a function of the two bias settings. These are a compromise between large signal handling, gain, noise, and frequency response. For these measurements, the transmitter bias (at IC1 pin 7) was set to -7.5V, and the receiver bias was set to 7V (at O1 pin 4) You may need to alter these to achieve best performance, depending on the individual optocouplers you use.
With a sine wave input of 1VRMS at 1kHz, the harmonic distortion at the output was examined using the FFT function of the Rigol DS1052E oscilloscope. The 2nd harmonic distortion appears to be around -40/-45dB. While not representing 'hi-fi' performance, this is quite adequate for the application.
Transfer function (x=input 0.5V/div, y=output 1V/div)
Frequency spectrum of output signal
The linearity was confirmed by plotting the output against the input. This showed a straight line for a 1VRMS signal, although some phase shift was evident. (A proper analogue scope is probably best for this application, as the X-Y mode on cheap DSOs can be rather miserable.) Significant clipping was not evident until the input amplitude exceeded 4VRMS
The frequency was then increased, and the -3dB frequency was found to be around 21kHz. However, further testing showed that this could be significantly lower at other bias settings. Overall gain was slightly under 2, again, this depends on the bias setting. With no signal input, there was around 20mVpp of noise on the output. This corresponds to a signal-to-noise ratio of around 55dB for full output. This is not as good as that of the other components of the system, but it should be adequate provided that the levels are matched reasonably well. It was subsequently found that this noise was inaudible above the background noise in the workshop at typical listening levels. Significantly, there was no sign of mains hum on the output, which justified this choice of circuit.
The circuit presented here provides a simple solution for transferring an audio signal across a galvanic isolation barrier. This is useful for interconnecting audio equipment in different rooms of a building, as it avoids introducing hum and noise onto the signal that might result from earth loops if the equipment were directly connected.
Another approach to this problem is to use an FM transmitter to send the signal. I had been using this approach originally, and while it was easy to set up, it suffered from frequent signal dropouts, which quickly became very annoying! The FM stereo multiplexing process can also result in added noise. The wired solution does not have these drawbacks. However, I really need to get the wiring properly installed in the roof, rather than having it running across the floor, as it is now!
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.